Tissue samples of 213 patients with epithelial ovarian tumors were included in our western blot analysis, including benign cystadenomas (n=5), borderline tumors (n=19), invasive primary carcinomas (n=178) and recurrent carcinomas (n=11). The cohort characteristics are summarized in Table I. CDH11 mRNA expression was analyzed in 51 samples, including 6 cystadenomas, 20 borderline tumors and 25 invasive primary carcinomas. CDH11 immunohistochemistry was performed on 3 cystadenomas, 12 borderline ovarian tumors and 8 invasive carcinomas. Surgery was performed at the University Medical Centre Hamburg-Eppendorf between 1994 and 2012, or at the Albertinen Hospital in Hamburg, between 2013 and 2014. All the patients provided written informed consent for examining their tissue samples and reviewing their medical records, according to our Investigational Review Board and Ethics Committee guidelines. A detailed database including clinicopathological factors, histological classifications and therapeutic procedures was generated. The clinical outcomes of all the patients were monitored from the date of surgery until December, 2013.

The tissue samples were collected intraoperatively and were immediately cryoconserved at −80°C. In order to assure a tumor cell content of ≥70%, every sample was assessed on cryo-cut sections stained with haematoxylin and eosin. If necessary, stromal parts were removed. Approximately 100 mg of tumor tissue were used for protein extraction, which was performed as previously described (13).

Western blot analysis was generally performed as previously described (13). Equal amounts of protein (20 µg) from each sample were loaded per well and equal loading was verified by immunoblotting with anti-GAPDH antibody (FL335; Santa Cruz, Heidelberg, Germany; 1:5,000). Proteins from a tumor with known moderate CDH11 expression served as control on each gel. Following electrophoresis and blotting to polyvinylidene difluoride membranes, the membranes were stored at 4°C in blocking solution. The rabbit anti-OB-cadherin antibody (P707; Cell Signaling Technology Inc., Danvers, MA, USA) was added to this blocking solution to a final concentration of 0.064 µg/ml and incubated overnight. Peroxidase-conjugated anti-rabbit IgG (sc-2054; dilution, 1:8,000; Santa Cruz) served as the secondary antibody and was visualized by chemiluminescence reagent (Super Signal West Pico kit; Pierce, Rockford, IL, USA) on medical X-ray films (Fujifilm Europe, Düsseldorf, Germany). The band intensities were quantified by densitometry (Imaging Densitometer GS-700; BioRad, Munich, Germany) and calculated as % intensity of the control tumor sample (CT).

RNA extraction from frozen ovarian tumor tissue and subsequent quality analysis were performed as previously described (14). RNA (5 µg) was reverse-transcribed using the Maxima First Strand cDNA Synthesis kit (Thermo Fisher Scientific, Pinneberg, Germany) and the obtained cDNA was diluted (1:5) for further RT-qPCR analysis. The following primers were used for amplification of the CDH11 sequences and the housekeeping gene GAPDH: CDH11: forward, 5′-CCCAGTACACGTTGATGCCT-3′ and reverse, 5′-GACGTTCCCACATTGGACCT-3′; GAPDH: forward, 5′-GTCAGTGGTGGACCTGACCT-3′ and reverse, 5′-TGCTGTAGCCAAATTCGTTG-3′. RT-qPCR was performed using the capillary-based Light Cycler (Roche, Basel, Switzerland) and the SYBR Premix Ex Taq (Takara Bio, Saint-Germain-en-Laye, France) with 1 µl cDNA. The samples were analyzed in duplicates and the results were averaged. CDH11 expression was normalized to the reference gene GAPDH and the relative expression of CDH11 in each sample was compared with the expression in one borderline sample, which was used as control (fold change = 1) based on the ΔΔCt method.

For CDH11 immunohistochemistry, paraffin-embedded tissue sections of 4 µm were deparafinized and incubated overnight at 4°C with the goat polyclonal CDH11 antibody (cat. no. AF1790; R&D Systems, Abingdon, UK), diluted to 1:150 in antibody diluent (Dako, Glostrup, Denmark) without prior antigen retrieval steps. For detection, the washed slides were incubated with biotin-labelled anti-mouse immunoglobulin (IgG), preformed ABC-Complex (Vectastain; Vector Laboratories, Burlingame, CA, USA) and DAB-substrate kit (Vectastain; Vector Laboratories). All the slides were counterstained with haematoxylin. For negative controls, normal rabbit IgG (Dako) was used instead of primary antibody. The endothelial antigen CD31 was visualized using the monoclonal mouse antibody anti-CD31 (cat. no. M0823; Dako, Hamburg, Germany), diluted 1:60. Images were captured using an AxioVision 40 microscope (Carl Zeiss Imaging Solutions, Munich, Germany) and photoshop software (Adobe Systems Inc., San Jose, CA, USA).

In a first approach, CDH11 mRNA expression was investigated in 6 cystadenomas/cystadenonofibromas, 20 borderline tumors and 25 invasive ovarian carcinomas. Interestingly, there was a distinct difference between these tumor types, with significantly lower CDH11 expression in carcinomas compared to borderline tumors (P<0.00001) or cystadenomas (P<0.00001). There was no statistically significant difference between benign or borderline tumors (P=0.14, Fig. 1A). Among invasive carcinomas, CDH11 expression was lower in poorly differentiated tumors (grade 3) compared with that in highly or moderately differentiated carcinomas (grade 1–2, P=0.004, Fig. 1B). No significant correlation to age, stage, nodal involvement or histological tumor type was observed (data not shown).

By western blot analysis, CDH11 was detected in almost all 213 tissue samples in varying amounts (Fig. 2A). Compared to the CT included in all the experiments, the protein expression levels ranged from 0.03 to 400% (mean, 90.5%; median, 72.0%). Regarding the tumor types, the mean CDH11 protein expression decreased with increasing malignancy, being 147% in benign tumors, 124% in borderline tumors and 86% in carcinomas (Fig. 2B). The difference between invasive cancer and borderline tumors was statistically significant (P=0.045). Recurrent tumors exhibited higher CDH11 protein levels (96%) compared with primary carcinomas (86%), but this difference did not reach statistical significance.

In 39 tumor samples representing all tumor types, CDH11 expression was analyzed at the protein and mRNA levels. By the Pearsons correlation test, there was no significant correlation between mRNA and protein data in this sub-cohort (r=0.280; P=0.084).

Based on CDH11 protein expression values, the cohort of 178 primary carcinomas was first divided into 4 quartiles of equal size with low, moderate, strong and very strong CDH11 expression. Comparing these 4 groups by Chi-square tests or using the median value as cut-off, there was no significant association with grade, age, International Federation of Gynecology and Obstetrics stage or lymph node involvement (data not shown). Yet, if the first quartile was compared with the upper 3 quartiles, low CDH11 expression was found to be significantly correlated with early stage and negative lymph node status (Fig. 2C and D). Regarding the histological subgroups, there was a non-significant tendency showing lower CDH11 expression in endometrioid carcinomas compared with serous tumors (data not shown). No other significant associations with clinical or histological parameters were observed in this group. By Kaplan-Meier analysis and log-rank test using different CDH11 cut-off values, no significant correlation with recurrence-free or overall survival was identified (data not shown).

Since CDH11 expression had been reported not only in tumor cells, but also in stromal or endothelial cells, we analyzed representative tumors from each group (8 invasive carcinomas, 12 borderline tumors and 3 cystadenomas) for CDH11 expression by immunohistochemistry. Strikingly, invasive carcinoma cells were either CDH11-negative or displayed only weak nuclear or cytoplasmic CDH11 expression. We did not detect any membranous CDH11 staining in tumor cells. By contrast, strong CDH11 staining was observed in the neighboring vessel walls and stromal matrix (Fig. 3). The comparison with immunohistochemical detection of the endothelial marker CD31 in parallel sections demonstrated CDH11 expression in endothelia and vascular smooth muscle cells around the blood vessels (Fig. 3C). Isotypic controls were always negative (data not shown).